spaCy/spacy/kb.pyx

248 lines
8.3 KiB
Cython
Raw Normal View History

2019-04-10 17:06:09 +03:00
# cython: infer_types=True
2019-03-18 19:27:51 +03:00
# cython: profile=True
# coding: utf8
from collections import OrderedDict
from cpython.exc cimport PyErr_CheckSignals
from spacy import util
2019-03-22 18:55:05 +03:00
from spacy.errors import Errors, Warnings, user_warning
from cpython.mem cimport PyMem_Malloc
from cpython.exc cimport PyErr_SetFromErrno
from libc.stdio cimport FILE, fopen, fclose, fread, fwrite, feof, fseek
from libc.stdint cimport int32_t, int64_t
from libc.stdlib cimport qsort
from .typedefs cimport hash_t
from os import path
cdef class Candidate:
def __init__(self, KnowledgeBase kb, entity_hash, alias_hash, prior_prob):
self.kb = kb
self.entity_hash = entity_hash
2019-03-21 14:31:02 +03:00
self.alias_hash = alias_hash
self.prior_prob = prior_prob
@property
def entity(self):
"""RETURNS (uint64): hash of the entity's KB ID/name"""
return self.entity_hash
2019-03-21 20:20:57 +03:00
@property
def entity_(self):
"""RETURNS (unicode): ID/name of this entity in the KB"""
2019-04-10 17:06:09 +03:00
return self.kb.vocab.strings[self.entity_hash]
@property
def alias(self):
"""RETURNS (uint64): hash of the alias"""
return self.alias_hash
@property
def alias_(self):
2019-03-21 20:20:57 +03:00
"""RETURNS (unicode): ID of the original alias"""
2019-04-10 17:06:09 +03:00
return self.kb.vocab.strings[self.alias_hash]
2019-03-21 20:20:57 +03:00
@property
def prior_prob(self):
return self.prior_prob
cdef class KnowledgeBase:
2019-03-18 19:27:51 +03:00
2019-03-22 01:17:25 +03:00
def __init__(self, Vocab vocab):
self.vocab = vocab
2019-03-18 19:27:51 +03:00
self.mem = Pool()
self._entry_index = PreshMap()
self._alias_index = PreshMap()
self.vocab.strings.add("")
self._create_empty_vectors(dummy_hash=self.vocab.strings[""])
2019-03-18 19:27:51 +03:00
def __len__(self):
2019-03-19 17:51:56 +03:00
return self.get_size_entities()
def get_size_entities(self):
return self._entries.size() - 1 # not counting dummy element on index 0
def get_entity_strings(self):
return [self.vocab.strings[x] for x in self._entry_index][1:] # removing the dummy element on index 0
2019-03-19 17:51:56 +03:00
def get_size_aliases(self):
return self._aliases_table.size() - 1 # not counting dummy element on index
def get_alias_strings(self):
return [self.vocab.strings[x] for x in self._alias_index][1:] # removing the dummy element on index 0
2019-03-19 17:51:56 +03:00
def add_entity(self, unicode entity, float prob=0.5, vectors=None, features=None):
"""
2019-04-10 17:06:09 +03:00
Add an entity to the KB, optionally specifying its log probability based on corpus frequency
Return the hash of the entity ID/name at the end
"""
cdef hash_t entity_hash = self.vocab.strings.add(entity)
2019-03-18 19:27:51 +03:00
# Return if this entity was added before
if entity_hash in self._entry_index:
user_warning(Warnings.W018.format(entity=entity))
return
2019-03-18 12:31:01 +03:00
cdef int32_t dummy_value = 342
new_index = self.c_add_entity(entity_hash=entity_hash, prob=prob,
vector_rows=&dummy_value, feats_row=dummy_value)
self._entry_index[entity_hash] = new_index
2019-03-18 12:31:01 +03:00
# TODO self._vectors_table.get_pointer(vectors),
2019-03-18 14:38:40 +03:00
# self._features_table.get(features))
return entity_hash
2019-03-18 19:27:51 +03:00
def add_alias(self, unicode alias, entities, probabilities):
"""
For a given alias, add its potential entities and prior probabilies to the KB.
Return the alias_hash at the end
"""
# Throw an error if the length of entities and probabilities are not the same
if not len(entities) == len(probabilities):
2019-03-22 18:55:05 +03:00
raise ValueError(Errors.E132.format(alias=alias,
entities_length=len(entities),
probabilities_length=len(probabilities)))
# Throw an error if the probabilities sum up to more than 1
prob_sum = sum(probabilities)
if prob_sum > 1:
2019-03-22 18:55:05 +03:00
raise ValueError(Errors.E133.format(alias=alias, sum=prob_sum))
2019-03-22 01:17:25 +03:00
cdef hash_t alias_hash = self.vocab.strings.add(alias)
# Return if this alias was added before
if alias_hash in self._alias_index:
2019-03-22 18:55:05 +03:00
user_warning(Warnings.W017.format(alias=alias))
return
2019-03-19 18:15:38 +03:00
cdef vector[int64_t] entry_indices
cdef vector[float] probs
for entity, prob in zip(entities, probabilities):
entity_hash = self.vocab.strings[entity]
if not entity_hash in self._entry_index:
2019-03-22 18:55:05 +03:00
raise ValueError(Errors.E134.format(alias=alias, entity=entity))
entry_index = <int64_t>self._entry_index.get(entity_hash)
2019-03-19 18:15:38 +03:00
entry_indices.push_back(int(entry_index))
probs.push_back(float(prob))
2019-03-18 14:38:40 +03:00
new_index = self.c_add_aliases(alias_hash=alias_hash, entry_indices=entry_indices, probs=probs)
self._alias_index[alias_hash] = new_index
2019-03-18 19:27:51 +03:00
return alias_hash
2019-03-18 19:50:01 +03:00
def get_candidates(self, unicode alias):
""" TODO: where to put this functionality ?"""
2019-03-22 01:17:25 +03:00
cdef hash_t alias_hash = self.vocab.strings[alias]
2019-03-19 17:51:56 +03:00
alias_index = <int64_t>self._alias_index.get(alias_hash)
alias_entry = self._aliases_table[alias_index]
return [Candidate(kb=self,
entity_hash=self._entries[entry_index].entity_hash,
2019-03-21 14:31:02 +03:00
alias_hash=alias_hash,
prior_prob=prob)
for (entry_index, prob) in zip(alias_entry.entry_indices, alias_entry.probs)
if entry_index != 0]
def dump(self, loc):
# TODO: actually dump the data in this KB :-)
cdef int64_t entry_id = 32
self.vocab.strings.add("Q342")
cdef hash_t entity_hash = self.vocab.strings["Q342"]
cdef float prob = 0.333
cdef Writer writer = Writer(loc)
writer.write(entry_id, entity_hash, prob)
writer.close()
def load(self, loc):
cdef int64_t entry_id
cdef hash_t entity_hash
cdef float prob
cdef Reader reader = Reader(loc)
reader.read(self.mem, &entry_id, &entity_hash, &prob)
cdef _EntryC entry
entry.entity_hash = entity_hash
entry.prob = prob
# TODO
cdef int32_t dummy_value = 342
entry.vector_rows = &dummy_value
entry.feats_row = dummy_value
cdef class Writer:
def __init__(self, object loc):
if path.exists(loc):
assert not path.isdir(loc), "%s is directory." % loc
cdef bytes bytes_loc = loc.encode('utf8') if type(loc) == unicode else loc
self._fp = fopen(<char*>bytes_loc, 'wb')
assert self._fp != NULL
fseek(self._fp, 0, 0)
def close(self):
cdef size_t status = fclose(self._fp)
assert status == 0
cdef int write(self, int64_t entry_id, hash_t entry_hash, float entry_prob) except -1:
cdef int i = 0
# TODO: feats_rows and vector rows
_write(&entry_id, sizeof(entry_id), self._fp)
_write(&entry_hash, sizeof(entry_hash), self._fp)
_write(&entry_prob, sizeof(entry_prob), self._fp)
cdef int _write(void* value, size_t size, FILE* fp) except -1:
status = fwrite(value, size, 1, fp)
assert status == 1, status
cdef class Reader:
def __init__(self, object loc):
assert path.exists(loc)
assert not path.isdir(loc)
cdef bytes bytes_loc = loc.encode('utf8') if type(loc) == unicode else loc
self._fp = fopen(<char*>bytes_loc, 'rb')
if not self._fp:
PyErr_SetFromErrno(IOError)
status = fseek(self._fp, 0, 0) # this can be 0 if there is no header
def __dealloc__(self):
fclose(self._fp)
cdef int read(self, Pool mem, int64_t* entry_id, hash_t* entity_hash, float* prob) except -1:
status = fread(entry_id, sizeof(entry_id), 1, self._fp)
if status < 1:
if feof(self._fp):
return 0 # end of file
raise IOError("error reading entry ID from input file")
#status = fread(&entity_hash, sizeof(entity_hash), 1, self._fp)
status = fread(entity_hash, sizeof(entity_hash), 1, self._fp)
if status < 1:
if feof(self._fp):
return 0 # end of file
raise IOError("error reading entity hash from input file")
status = fread(prob, sizeof(prob), 1, self._fp)
if status < 1:
if feof(self._fp):
return 0 # end of file
raise IOError("error reading entity prob from input file")